FitCheck

Chest Fly Analysis - 0:03
Chest Fly Analysis - 0:06
Chest Fly Analysis - 0:13
Chest Fly Analysis - 0:21
Chest Fly Analysis

Inspiration

All 3 of us have been working out but with so many different exercises and workout routines on the internet, it gets really overwhelming and learning new exercises still may not feel right even with years of experience. We've all felt when we see an amazing exercise online, try to replicate it, but it doesn't feel right or may be working something else entirely. So, why not use Gemini to give us quick and accurate feedback on our form?

What it does

FitCheck allows you to upload a video of yourself doing an exercise. After some processing, FitCheck will return detailed feedback on how to improve and even specify the areas using computer vision, where your troublesome areas are.

How we built it

We built this with a variety of tech. The bulk of our processing is handled by Gemini, but we also have detailed logic and support for video uploading using S3, video parsing, and computer vision to point out the exact body parts on the video and highlight them.

Challenges we ran into

Parsing videos into 1 frame per second, as limited by Gemini, was definitely difficult to pull off. Furthermore, actually connecting these moving components within the given time was difficult as well. From React all the way to supporting Gemini using FastAPI, we had to learn and execute in a limited time.

Accomplishments that we're proud of

We are proud of successfully integrating various technologies, such as Gemini, S3, video parsing, and computer vision, to create a cohesive and functional application. Despite the challenges we faced, we managed to build a working prototype that demonstrates the potential of using AI to provide personalized feedback on exercise form.

What we learned

Throughout the development process, we gained valuable experience in working with different technologies and integrating them seamlessly. We learned how to leverage Gemini's capabilities to analyze exercise form, utilize S3 for efficient video storage and retrieval, and apply computer vision techniques to highlight specific body parts in the video. Additionally, we improved our skills in React development and FastAPI implementation, enabling us to build a user-friendly interface and a robust backend system.

What's next for FitCheck

Hopefully we can switch off of React and make a mobile application for this! I think it'd be super convenient and quick to use for anyone trying to get into working out.